[1]胡今鸣,胡啸峰,石磊,等.基于强化学习的超高层建筑非法入侵情景推演方法[J].智能系统学报,2025,20(4):958-968.[doi:10.11992/tis.202408002]
HU Jinming,HU Xiaofeng,SHI Lei,et al.Method of unauthorized intrusion scenario simulation in super high-rise building based on reinforcement learning[J].CAAI Transactions on Intelligent Systems,2025,20(4):958-968.[doi:10.11992/tis.202408002]
点击复制
HU Jinming,HU Xiaofeng,SHI Lei,et al.Method of unauthorized intrusion scenario simulation in super high-rise building based on reinforcement learning[J].CAAI Transactions on Intelligent Systems,2025,20(4):958-968.[doi:10.11992/tis.202408002]
基于强化学习的超高层建筑非法入侵情景推演方法
《智能系统学报》[ISSN 1673-4785/CN 23-1538/TP] 卷:
20
期数:
2025年第4期
页码:
958-968
栏目:
学术论文—机器学习
出版日期:
2025-08-05
- Title:
- Method of unauthorized intrusion scenario simulation in super high-rise building based on reinforcement learning
- Keywords:
- unauthorized intrusion; scenario simulation; super high-rise building; reinforcement learning; Bayesian network; security system; SARSA model; nonlinear regression
- 分类号:
- TP18; X937
- 文献标志码:
- 2025-1-15
- 摘要:
- 为计算超高层建筑潜在非法入侵者的“最优”入侵路径，本文提出了一种基于强化学习的情景推演方法。该方法将建筑公共走廊抽象为拓扑结构，利用贝叶斯网络计算入侵者通过每个拓扑节点的概率，结合强化学习算法获得外部人员的最优入侵路径，为超高层建筑非法入侵的高效防范提供精准依据。为验证方法的有效性，以北京市CBD地区某超高层建筑为例，将入侵终点设置为顶层，设计了3种不同的入侵情景。情景推演结果表明：在初始状态下（未进行任何优化措施），SARSA模型的训练性能最佳。优化安防系统后发现，在建筑内的层间节点增加安防系统投入最有效。该优化情景下，安防系统投入与风险值的非线性拟合结果显示，随着安防系统投入的增加，入侵风险显著降低。
- Abstract:
- To calculate the “optimal” intrusion path of potential illegal intruders in super high-rise buildings, a scenario simulation method based on reinforcement learning is proposed in the paper. This method provides a precise basis for efficiently preventing illegal access in super high-rise buildings by abstracting the buildings’ public corridors into a topological structure, calculating the probability of an intruder passing through each node based on a Bayesian network, and exploring the optimal intrusion path by means of reinforcement learning algorithms. To validate this method, a super high-rise building in the CBD area of Beijing was taken as an example, where the intrusion endpoint was assumed as the top floor and three different intrusion scenarios were designed. Results reveal that the SARSA model has the best training performance in the initial state (without any optimization measures). After optimizing the security system, increasing security system investment at interfloor nodes within the building is the most effective. In this context, a nonlinear fit between security investment and risk values shows that as investment in a security prevent system increases, intrusion risk remarkably decreases.
- 参考文献/References:
-
[1] 中华人民共和国住房和城乡建设部. 安全防范工程通用规范: GB 55029—2022[S]. 北京: 中国计划出版社, 2022.
Ministry of Housing and Urban-Rural Development of the People’s Republic of China. General code of security engineering: GB 55029—2022[S]. Beijing: China Planning Press, 2022.
[2] HUANG He, HU Hao, XU Feng, et al. Skeleton-based automatic assessment and prediction of intrusion risk in construction hazardous areas[J]. Safety science, 2023, 164: 106150.
[3] LI Heng, DONG Shuang, SKITMORE M, et al. Intrusion warning and assessment method for site safety enhancement[J]. Safety science, 2016, 84: 97-107.
[4] ARSLAN M, CRUZ C, GINHAC D. Visualizing intrusions in dynamic building environments for worker safety[J]. Safety science, 2019, 120: 428-446.
[5] 王润芳, 陈增强, 刘忠信. 融合朴素贝叶斯方法的复杂网络链路预测[J]. 智能系统学报, 2019, 14(1): 99-107.
WANG Runfang, CHEN Zengqiang, LIU Zhongxin. Link prediction in complex networks with syncretic naive Bayes methods[J]. CAAI transactions on intelligent systems, 2019, 14(1): 99-107.
[6] GUO Kai, ZHANG Limao, WU Maozhi. Simulation-based multi-objective optimization towards proactive evacuation planning at metro stations[J]. Engineering applications of artificial intelligence, 2023, 120: 105858.
[7] 李冰, 杨薪玉, 王延锋. 轨道交通车站乘客集散系统Anylogic仿真优化[J]. 智能系统学报, 2020, 15(6): 1049-1057.
LI Bing, YANG Xinyu, WANG Yanfeng. Simulation and optimization of the passenger distribution system Anylogic in rail transit stations[J]. CAAI transactions on intelligent systems, 2020, 15(6): 1049-1057.
[8] HOSSEINI S, IVANOV D. Bayesian networks for supply chain risk, resilience and ripple effect analysis: a literature review[J]. Expert systems with applications, 2020, 161: 113649.
[9] ZHU Rongchen, HU Xiaofeng, BAI Yiping, et al. Risk analysis of terrorist attacks on LNG storage tanks at ports[J]. Safety science, 2021, 137: 105192.
[10] BIAN Tao, JIANG Zhongping. Reinforcement learning and adaptive optimal control for continuous-time nonlinear systems: a value iteration approach[J]. IEEE transactions on neural networks and learning systems, 2022, 33(7): 2781-2790.
[11] 于泽, 宁念文, 郑燕柳, 等. 深度强化学习驱动的智能交通信号控制策略综述[J]. 计算机科学, 2023, 50(4): 159-171.
YU Ze, NING Nianwen, ZHENG Yanliu, et al. Review of intelligent traffic signal control strategies driven by deep reinforcement learning[J]. Computer science, 2023, 50(4): 159-171.
[12] WU Yan, LUO Shixian, DENG Feiqi. Reinforcement learning for optimal control of linear impulsive systems with periodic impulses[J]. Neurocomputing, 2024, 585: 127569.
[13] 高玉钊, 聂一鸣. 基于值函数分解的多智能体深度强化学习方法研究综述[J]. 计算机科学, 2024, 51(S1): 34-42.
GAO Yuzhao, NIE Yiming. Review of multi-agent deep reinforcement learning method based on value function decomposition[J]. Computer science, 2024, 51(S1): 34-42.
[14] HU Jinming, HU Xiaofeng, KONG Feng, et al. Vulnerability analysis of super high-rise building security system based on Bayesian network and digital twin technology[J]. Process safety and environmental protection, 2024, 187: 1047-1061.
[15] ASGHARNIA A, SCHWARTZ H, ATIA M. Multi-objective fuzzy Q-learning to solve continuous state-action problems[J]. Neurocomputing, 2023, 516: 115-132.
[16] KIUMARSI B, ALQAUDI B, MODARES H, et al. Optimal control using adaptive resonance theory and Q-learning[J]. Neurocomputing, 2019, 361: 119-125.
[17] GARí Y, PACINI E, ROBINO L, et al. Online RL-based cloud autoscaling for scientific workflows: Evaluation of Q-Learning and SARSA[J]. Future generation computer systems, 2024, 157: 573-586.
[18] YANG Xu, LIU Pei, LIU Fang, et al. A DOD-SOH balancing control method for dynamic reconfigurable battery systems based on DQN algorithm[J]. Frontiers in energy research, 2023, 11: 1333147.
[19] WU Peiliang, ZHANG Yan, LI Yao, et al. A robot pick and place skill learning method based on maximum entropy and DDQN algorithm[J]. Journal of physics: conference series, 2022, 2203(1): 012063.
[20] SENTHIL KUMAR S, ALZABEN N, SRIDEVI A, et al. Improving quality of service (QoS) in wireless multimedia sensor networks using epsilon greedy strategy[J]. Measurement science review, 2024, 24(3): 113-117.
[21] MILI K, BENGANA I, OUASSAF S, et al. Testing the co-integration relationship between auto insurance premiums and risk compensation amount[J]. Computers in human behavior reports, 2024, 13: 100377.
[22] HOU Miaomiao, HU Xiaofeng, CAI Jitao, et al. An integrated graph model for spatial-temporal urban crime prediction based on attention mechanism[J]. ISPRS international journal of geo-information, 2022, 11(5): 294.
[23] WANG Lina, XU Mengjie, ZHANG Ying. An intelligent decision algorithm for a greenhouse system based on a rough set and D-S evidence theory[J]. IAENG international journal of applied mathematics, 2024, 54(6): 1240-1250.
[24] 秦荣水, 石晨晨, 陈超, 等. 基于模糊贝叶斯网络的城市商业综合体火灾风险分析[J]. 中国安全科学学报, 2023, 33(12): 176-182.
QIN Rongshui, SHI Chenchen, CHEN Chao, et al. Risk analysis on fire accident of urban commercial complex based on fuzzy Bayesian network[J]. China safety science journal, 2023, 33(12): 176-182.
[25] COPPA E, IZZILLO A. Testing concolic execution through consistency checks[J]. Journal of systems and software, 2024, 211: 112001.
备注/Memo
收稿日期:2024-8-3。
基金项目:中国人民公安大学拔尖创新人才培养研究生科研创新重点项目(2024yjsky009)；国家自然科学基金项目(72174203)；中国人民公安大学安全防范工程双一流专项(2023SYL08).
作者简介:胡今鸣，硕士研究生，主要研究方向为强化学习、社会公共安全风险评估。发表学术论文4篇。E-mail：hujinming2024@163.com。;胡啸峰，副教授，博士，主要研究方向为人工智能、社会公共安全风险评估。主持国家自然科学基金项目2项，发表学术论文60余篇。E-mail：huxiaofeng@ppsuc.edu.cn。;石磊，助理研究员，博士，中国人工智能学会智能服务专委会委员，主要研究方向为智能信息处理、大数据分析与挖掘、社交网络搜索及人工智能。发表学术论文20余篇。E-mail：leiky_shi@cuc.edu.cn。
通讯作者:胡啸峰. E-mail：huxiaofeng@ppsuc.edu.cn
更新日期/Last Update:
1900-01-01